- AWS unveils new AI drug discovery tool
- Amazon Bio Discovery removes the technical barriers to high computational AI experiments
- The tool can significantly reduce drug testing time
A new AI-powered drug discovery tool has been launched by Amazon Web Services (AWS).
The Amazon Bio Discovery tool helps researchers accelerate the discovery of new drugs by allowing scientists to run complex computational workloads without the need for technical expertise.
Amazon’s cloud platform touts the tools as being able to reduce the timescale of an antibody design workflow from 12 months to just weeks.
The article continues below
AI accelerates drug discovery
Amazon Bio Discovery provides a catalog of basic models specialized for drug discovery, with the option for researchers to upload models from third parties. Of course, the tool would not be complete without an AI agent that can guide users through choosing the right models and parameters for their research.
When the experiment is ready to start, the AI agent begins searching through data sources and basic biological factors—and it even provides references and scientific justifications for its predictions and suggestions.
The tool then filters the results down to the top selection of results, which can then be sent to one of Amazon’s integrated lab partners for synthesis and testing without the need for a manual handoff that can cause delays. The results from laboratory testing are then automatically fed back to Amazon Bio Discovery for further analysis.
The continuous back-and-forth feedback between the integrated laboratories and researchers allows for rapid fine-tuning of results, speeding up the time between design, testing and synthesis.
In collaborative testing with Memorial Sloan Kettering Cancer Center, Amazon Bio Discovery helped narrow a selection of 300,000 antibody candidates to the top 100,000 and sent them to test “in weeks versus up to a year using traditional design methods.”
AWS also partnered with the Gray Lab at the Johns Hopkins Whiting School of Engineering to produce the ‘Antibody Developability Benchmark’ – the “largest and most diverse” antibody dataset designed to help evaluate AI-guided antibody design.
Luca Giancardo, an applied scientist at Amazon Web Services said: “This data set will allow researchers to confidently answer ‘Which model is better suited for our purposes?’. Today there are many computational models coming out that are mostly evaluated on either proprietary data or public data sets that are not representative of antibody heterogeneity – whether it is much better or not is difficult to decide.”
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews and opinions in your feeds. Be sure to click the Follow button!
And of course you can too follow TechRadar on TikTok for news, reviews, video unboxings, and get regular updates from us on WhatsApp also.
